Application And Implement Of Face Recognition And Human Detection Technologies In Unattended Substation

Current researches on substation demonstrate that they are tending to be non-attended. Safety requirements in the station are simultaneously becoming higher.Although the four remote control technology can monitor the station equipment and environment, they don't take precautions against the peripheral and access area.Remote watching makes the comprehensive monitoring of the substation in bot interior and exterior, however, it is mainly used to record and save the working conditions. In a word, the overall security level can not meet the requirements in the station. Therefore, in this paper, we intend to utilize the face recognition and human detection technologies to enrich the function of remote watching and ulteriorly enhance the substation's security level. Our specific work are as follows:The video monitoring image in substation will be affected by the illumination,noise and other factors. We wield gray scale transformation, noise filtering, image enhancement and scale normalization for image preprocessing.Face recognition algorithms in the application of substation access control,mainly related to the human-face feature extraction, classification and recognition after feature extraction. According to the environmental characteristics of the substation, the paper analyzes and implements the feature extraction and classification algorithms, and we successfully find two human-face recognition algorithms which are suitable for the working environment of substation. Compared with PCA and LDA,LBP algorithm has a higher resistance to light and a better recognition rate. In order to reduce the influence of noise and improve the recognition rate, we adopt the PCA algorithm to reduce the dimension and enhance the LBP algorithm. In the case of a small number of samples, we compare the recognition effects of KNN, SVM and BP neural network. The recognition rate of KNN algorithm gets higher and the SVM algorithm behaves more efficiently. Finally, based on the above results, face recognition methods we employ in the substation to improve access control are the combination of improved LBP and KNN-SVM.The application the human detection algorithms that monitor abnormal activities in the working area of the substation, mainly related to the feature extraction aims at human bodies, as well as the classification. On account of the functions of the human detection in the substation area and these environmental characteristics of the substation, we choose the better detection algorithm from three aspects which are feature extraction, classifiers and detection modes. On the issues of feature selection,we compare the detection results of HOG features from head to shoulders to the whole body. When the body is blocked and posture changes, better detection effects are on human head and shoulders. Then, two classifiers which are SVM and cascade Adaboost are compared. The results show that cascade Adaboost performs better. In the case of detection method, we use the method of image gold tower to search for multi-scale spatial detection, which is more in line with the actual. Eventually,multi-scale human detection method that human head and shoulders HOG combines cascade Adaboost is selected for substation's monitoring and early warning.In the end, based on face recognition and human detection, we design and implement a substation monitoring system, which has the functions of face recognition in the access control and human detection in various areas. We achieve the function of face recognition and human detection, which response to the results of its response by operating data tables.